Magic-League.com Forum Index Magic-League.com
Forums of Magic-League: Free Online tcg playing; casual or tournament play.
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Feedback: MetagameBreakdown Tool



 
Reply to topic    Magic-League.com Forum Index -> Magic-League Tournaments
Author Message
Burton911



Joined: 09 Jun 2007
Posts: 172

PostPosted: Tue Aug 30, 2011 10:38 am    Post subject: Feedback: MetagameBreakdown Tool Reply with quote

You might have noticed that ive posted Meta Breakdown in the comments of trials, over the last couple of months, and today even for a limited trial in the limited section of the forum.

I open this thread, so that there is a place where you can tell any ideas you have which might improve the programm.

List of things I may going to implement:
*) compare the last current trial to the last trial in this format (im definatly doing this one)
*) gahtering empyrical data

[DISCLAIMER] The Feedback part ends here, the last part is kind of "tl;dr like" but might improve the programm ^^ [/DISCLAIMER]


And i also have a concret request to anyone whoe likes to help:
I want to improve the process which assigns the archtypes to the decks.
Right now it works the following way:

There are archtype files, which are structered as follwing:
//Name: {name here}
//DATE: {dd.mm.yyyy}
cardname
cardname
!cardname

(For example this is the archtype list as used for modern right now)

I added the date in the beginning but its not used in the programm right now.
The cards with a ! before them are cards which are excluded by this archtype.
So a deck is of a certain archtype if it contains all cards without a ! and not contains any card with an ! at the beginning.

Overall this works fine, tough some archtypes are different only by a few cards, which makes them hard to determine. Thats why im asking for ideas for new Filter criterias, or might even a completly new process of filtering the archtypes.
Right now i tought of the following ways:
new criterias:
*) check for certain amounts: This would check if there are lets say 4 copies of a card in the deck, so if there are less copies the deck cant be of this archtype
*) check for cardgroups: This means that there are entries in the archtype files that look like the following: #2#card1/card2/card3
In this example the deck has to contain 2 of the 3 cards to be categorized in this archtype.

Completly new approches would be the following two:
*) best fitting core: There would be a core of cards for each archtype, and a deck would be assigned the archtype which core fits best. I only list this one as an example cause i think its worse then the current approach.

*) best fit in archtype cardpool: This one would assign a deck to the archtype in which pool it fits best.
The Problem of this one is that still similar archtypes are kind of hard to detect, tough i think that it could be better than the current approach.

The problem of the current process is that the filter criterias are quite strong. The best example why this is bad are the 3 Color Aggro Control decks of Legacy. Cause there the filter is too weak, which means that soem archtypes overlap or its too strong, which means that decks that should be classified in this archtype are left out.
This problem might could be solved by the "best fit in archtype cardpool"-approach since this one assigns the archtype two a deck where it fits best.

I hope its possible to understand what im trying to say, but im no native english speaker :/
If you want to tell me something on IRC ma nick there is: STEB

cu
Burton911
Back to top
Burton911



Joined: 09 Jun 2007
Posts: 172

PostPosted: Wed Aug 31, 2011 6:11 pm    Post subject: Reply with quote

as i said in the comments of todays 1.5 trial i need some help with the archtypes of legacy mainly cause i dont know the names of many Razz

The following link guides you to the current file used for determining archtypes: click me hard

What i need is at least the a list with the names of the imporant archtypes. Its not necessary that they appear in the file, since i can enter them by hand now easily.
Especially the names of the different versions of Goyf Aggro Control decks would be nice, cause the last time when i often played Legacy there where only different incarnations of Counterbalance but now there are the bant decks, the natural order decks, Jace based decks, stoneforge based decks and a few less played ones.

thx for reading
Burton911


Last edited by Burton911 on Thu Sep 01, 2011 5:53 am; edited 1 time in total
Back to top
Dzy



Joined: 03 Sep 2004
Posts: 148

PostPosted: Thu Sep 01, 2011 5:20 am    Post subject: Reply with quote

Hey I don't have a lot of time at the moment, but I will comment on a couple of decks:

I would cut thoughtcast from Affinity, not all the lists play it, and the other cards will never be played outside of affinity.

I would cut Seismic Assault from Aggro Loam, not all lists play it, I would add Tranquil Thicket and Tarmogoyf.

I would cut Forest from Aluren, 99% will play at least 1 forest, but you never know and those two cards do define that archtype.

Does it check maindeck only? If so Doomsday is fine, but some Ad Nauseam decks play Doomsday in the side.

I would change forest to Solitary Confinement in Enchantress.

I would keep Goblins to just Goblin Lackey, there was a list doing well at SCG without Warchief (amazingly stupid but true)

Green Suns Order is not a deck, and this package can be found in either NO Bant and NO Rug, two completely different decks.

GWB Rockstyle deck is called Junk.

Karn Blue does not exist. (7 mana planeswalkers suck in legacy)

I would cut the Cursecatcher from Merfolk, it's a sucky creature and people who are good at magic will cut it from their deck.

Monobrown Metalworker is called MUD and usually doesn't play Trinisphere.

Next Level Wizards is not a deck, Venser isn't played. There are decks playing all the other cards, that's Faerie Mystic.

There is another deck playing Swamp and Reanimate, it's a bad Necrotic Ooze deck. By now every reanimator deck plays Jin-Gitaxias, so add that.

Island + Show and Tell is not only in Show and Tell decks, but also in Hive Mind. Is it possible to do an OR clausule in your program? If so, do an OR on either Progenitus or Blightsteel Colossus.

Not all Stax decks are white (however virtually unexistent atm)

TarmoJace is not a deck, and if played together they can be either a control deck with goyfs or Blue Zoo.

UW Jace Control should be distinct between SFM and no SFM.
Back to top
Burton911



Joined: 09 Jun 2007
Posts: 172

PostPosted: Thu Sep 01, 2011 11:13 am    Post subject: Reply with quote

Thanks for your effort.
I will adapt the file to you suggestions later.

To adress a few points you mentioned.
The idea behind the names like TarmoJace, Greens Suns Order and the likes is just to show that there are decks including that cards.
I just made them cause of the problem with all the very similar Tarmogoyf Control decks. Tough with the tool explained bellow and the possibility of excluding cards it should be possible to solve that problem.

@The OR clause: i more or less tought of a concept like that in form of the cardgroups i mentioned in the first post, and ive implemented that now. It works the following way.
The file has to include a line like this one:
#n#card1;card2;cardn;
n has to be of course <= the number of cards tough it only makes sense if it is smaller.
An ordinery OR clause would be: #1#card1;card2; now.
Back to top
Display posts from previous:   
Reply to topic    Magic-League.com Forum Index -> Magic-League Tournaments All times are GMT - 7 Hours
Page 1 of 1

 


Powered by phpBB © 2001, 2005 phpBB Group

All content on this page may not be reproduced without consent of Magic-League Directors.
Magic the Gathering is TM and copyright Wizards of the Coast, Inc, a subsidiary of Hasbro, Inc. All rights reserved.


About Us | Contact Us | Privacy Policy